Skip to content
Cogito AIEnterprise AI Operating System

The right model for every agent, with no lock-in

Combine a wide range of models with your enterprise data and controls, and choose the right model for each use case to manage inference cost.

  • Choose the right model for each task based on quality, latency and cost
  • Lower inference spend without sacrificing quality where it counts
  • Outputs grounded in your data, not generic training data
  • Consistent privacy and residency controls across all models
  • Adopt newer models without re-platforming
  • Leverage open-weight models where appropriate
  • For the highest level of security, a completely ring-fenced inference solution is available

How models work in Cogito AI

Cogito AI lets you match each use case to the right model on quality, latency and cost, so routine work runs on efficient models and premium models are used only where they pay off. Every model is grounded in your data, with enterprise controls over privacy, retention and access.

  • 01

    Flexible LLM Choice

    Choose which models are available to your users and agents, set a default, and switch whenever a better fit launches, with no lock-in.

  • 02

    Cost control

    Set a default model system-wide and for each agent, limit premium models to the users who actually need them, and analyze model cost by user, agent, use case and conversation.

  • 03

    Grounded in your data

    Combine any model with your connected knowledge and ontology so outputs reflect your business, not generic training data.

  • 04

    Controlled & private

    Enterprise controls over data flow, retention, residency and access apply regardless of which model is used.

  • 05

    Open-weight models

    Run models in a completely ring-fenced manner for the highest levels of security. Leverage open-weight models for the lowest cost.

  • 06

    Future-proof and improving

    Adopt new and better models as they emerge without re-architecting, and tune against your own success metrics so quality improves where it matters.

In the platform

See token use and cost by model, agent, user and channel, for the last 24 hours, 7 days or 30 days.

DashboardUsage and cost

Tokens consumed by model
Most capable model: 3.2M tokens, $12.80Balanced model: 4.1M tokens, $3.28Fast, low-cost model: 3.0M tokens, $0.45
  • Most capable model3.2M31%$12.80
  • Balanced model4.1M40%$3.28
  • Fast, low-cost model3.0M29%$0.45

Illustrative prices

Last 7 daysToken consumption by agent
AgentTokens Cost
1RFP Response6.1M$10.08
2Customer Relationship Intelligence2.6M$4.08
3Intelligent Document Search1.6M$2.37
Last 7 daysToken consumption by user
UserTokens Cost
1Jane Doe3.9M$6.26
2Sam Rivera3.1M$4.97
3Priya Shah1.8M$2.90
4Alex Chen0.9M$1.44
5Maria Lopez0.6M$0.96
Token consumption by channel
Platform: 8.7M tokens, $13.96Salesforce: 1.6M tokens, $2.57
  • Platform8.7M84%$13.96
  • Salesforce1.6M16%$2.57

Illustrative product view with sample data.

How it works

  1. 1

    Match model to task

    Choose the model that fits each use case's quality, latency and cost profile.

  2. 2

    Work from your data

    The model is grounded in your connected knowledge and ontology, so outputs reflect your business.

  3. 3

    Apply enterprise controls

    Data flow, retention, residency and access controls are enforced no matter which model runs.

  4. 4

    Measure and tune

    Evaluate outputs against your success metrics and adjust model choice or tuning to improve results.

Read moreAgents